Understanding How Ollama Stores Models 

I'm still waiting for some kind of out-take moment at the end of the video :-) Thanks for sharing.

Well done! Thank you.

Love you videos, can’t wait for the Windows version 😊

Good video. The sha256:... format filenames aren't supported on windows filesystems because of the colon (I noticed this when trying to backup to a drive that was shared with windows) so I wonder if this is one of the things that needs to be changed for the windows release.

you have any idea how to give the model context data similar. like how we can use python libraries to request from openai interact with it via code?

Windows ollama has now come out and I'm just wondering where it stores them when it comes to Windows?

At ollama, can I run a model through path and not by name? For example instead of 'ollama run ' and linux and with the python library 'ollama.run("model")" to put a path istead of a model.

Yes but more importantly how do I reuse the already downloaded models with other frameworks, because everyone is downloading between 5 to 30GB files + and it would be good if they'd all be sourced form the same image, is there a way to expose those ollama models for consumption with another engine? currently Jan doesn't recognise the way models are stored in ollama. *Edit and vice versa - I am thinking about vLLM and jan? Ps Ollama webui lite doesn't post in the chat, and Open-webui is broken I could not get the container to render the app, not matter how much I played with the compose yamls, but then again, I wanted GPU acceleration so I don't know if the deploy agrs are going to work without Swarm enabled: deploy: resources: reservations: devices: - driver: ${OLLAMA_GPU_DRIVER-nvidia} count: ${OLLAMA_GPU_COUNT-1} do you happen to know if those variables are recognisable as is system-wide or do they need to be i the usr/bin?